test: 100% coverage + CI coverage gate

[AkashBrowserStack]

What & why

This SDK had no coverage enforcement in CI and sat at ~95% line locally. This PR raises it to 100% line coverage with meaningful tests and wires a hard CI gate so coverage can't silently regress.

Every line is executed by a real unit test — no # pragma skips, no coverage-gaming. Tests exercise real logic: validation, error paths, and fallbacks.

Tests added

Area	Coverage added
PercyOnAutomate (was untested)	invalid driver, disabled, name/options validation, alt camelCase region-key normalisation + element-id extraction, swallow-and-return-None error path — new tests/test_percy_automate.py
percy/__init__.py	percySnapshot alias + percy_snapshot/percy_screenshot import fallbacks (the latter via a reload-based test) — new tests/test_init.py
cli_wrapper	old-CLI minor-version rejection, unsuccessful healthcheck, post_failed_event error swallow, post_poa_screenshots success + error
app_automate	remote-uploads-disabled fallback, begin() JSON parse, execute_percy_screenshot error, re-raise after failure notification
app_percy	sync / test_case / th_test_case_execution_id type guards
generic_provider	full-page falls back to single page
metadata	remote_url new-client path + old-client URL fallback
android_metadata	malformed viewportRect falls back to driver bars

Instrumentation / gate

• .coveragerc — source = percy, fail_under = 100, show_missing.
• coverage added to development.txt; make coverage reports and enforces the threshold.
• test.yml runs make coverage across the 3.9 / 3.11 / 3.12 matrix → red on any drop below 100%.

Verification

Ran 158 tests ... OK
TOTAL  775 stmts  0 miss  100%   (no pragma exclusions)
pylint: 10.00/10

🤖 Generated with Claude Code

[AkashBrowserStack]

Claude Code PR Review

PR: #208 • Head: db34cf5 • Reviewers: stack:code-reviewer

Summary

Adds unit tests across seven modules (app_automate, generic_provider, cli_wrapper, android_metadata, metadata, percy_automate, app_percy/init) covering previously-untested error, fallback, and edge-case paths, and introduces a CI coverage gate: a new .coveragerc with fail_under = 100, a coverage==7.* dev dependency, a Makefile coverage target switched to coverage report, and CI (test.yml) switched from make test to make coverage. No production/source code under percy/ is modified — this is a test-and-tooling-only PR.

Review Table

Priority	Category	Check	Status	Notes
High	Security	No hardcoded secrets or credentials	Pass	No secrets/tokens introduced. Test fixtures use dummy hub URLs and fake session ids only.
High	Security	Authentication/authorization checks present	N/A	No auth surface touched; tests-and-tooling-only PR.
High	Security	Input validation and sanitization	N/A	No new input-handling code; tests only assert existing validation (e.g. type checks in app_percy.screenshot).
High	Security	No IDOR — resource ownership validated	N/A	No resource-access code in scope.
High	Security	No SQL injection (parameterized queries)	N/A	No DB/SQL in this codebase.
High	Correctness	Logic is correct, handles edge cases	Pass	Test assertions verified against source. Version-gate (1.26.0 rejected, minor < 27), viewportRect-malformed fallback, remote_url new/old client, POA alt-key normalization, and screenshot failure-notification re-raise all match actual percy/ behavior.
High	Correctness	Error handling is explicit, no swallowed exceptions	Pass	Tests deliberately assert the swallow-and-return-None contracts (post_failed_event, percy_automate.screenshot, screenshot_begin) rather than hiding them; lru_cache is cleared in try/finally so failures don't leak cached state.
High	Correctness	No race conditions or concurrency issues	N/A	No concurrency introduced; tests are synchronous.
Medium	Testing	New code has corresponding tests	Pass	This PR is the tests; they target real, exercisable source paths (verified each assertion against the module under test).
Medium	Testing	Error paths and edge cases tested	Pass	Strong coverage of error/fallback branches: malformed viewportRect, disabled remote uploads, finalize-failure re-raise, healthcheck unsuccessful, old CLI version, POA exception swallow, package import-fallback ModuleNotFoundError.
Medium	Testing	Existing tests still pass (no regressions)	Pass	Additive; no existing test bodies changed. test_init reloads percy and restores it in finally; cli_wrapper tests clear the is_percy_enabled lru_cache around the call. (Low-sev isolation nit noted in Findings, non-blocking.)
Medium	Performance	No N+1 queries or unbounded data fetching	N/A	No data-fetch loops added; test code only.
Medium	Performance	Long-running tasks use background jobs	N/A	Not applicable to an SDK test suite.
Medium	Quality	Follows existing codebase patterns	Pass	Matches existing unittest + unittest.mock style, # pylint: disable=[...] headers, mock_methods fixtures, and PropertyMock patching used elsewhere in the suite.
Medium	Quality	Changes are focused (single concern)	Pass	Single concern: raise coverage to 100% and gate it in CI. The .python-version 3.10.3 -> 3.10 tweak is a minor adjacent change but consistent with the tooling focus.
Low	Quality	Meaningful names, no dead code	Pass	Test names are descriptive and intent-revealing; explanatory comments justify non-obvious branches. No dead code.
Low	Quality	Comments explain why, not what	Pass	Comments explain why a branch is exercised (e.g. "full page is only supported on App Automate -> warn and produce a single tile").
Low	Quality	No unnecessary dependencies added	Pass	Only coverage==7.* added, to a dev-only requirements file (development.txt); appropriate and from PyPI.

Findings

No Critical or High findings. Two non-blocking, low-severity observations:

• File: tests/test_android_metadata.py:71

• Severity: Low

• Reviewer: stack:code-reviewer

• Issue: The test mutates the module-level shared android_capabilities dict (self.mock_webdriver.capabilities['viewportRect'] = {'left': 5}) and the finally restores it to None rather than to its original loaded value. android_capabilities is loaded once at import and shared by reference across test classes, so restoring to None leaves a slightly different state than the original. In practice each test re-reads capabilities via setUp and no other test asserts on viewportRect, so there is no observed cross-test contamination — but it is a latent isolation fragility.

• Suggestion: Capture and restore the original value, e.g. orig = self.mock_webdriver.capabilities.get('viewportRect') before mutating and restore orig in finally; or deepcopy the caps in setUp.

• File: percy/providers/app_automate.py:27 (context surfaced by tests/test_app_automate.py:108 test_screenshot_reraises_after_failure_notification, not a defect in the test)

• Severity: Low (pre-existing source observation, out of this PR's scope)

• Reviewer: stack:code-reviewer

• Issue: In AppAutomate.screenshot, if super().screenshot() raises before percy_screenshot_url is assigned, the except handler references the then-unbound percy_screenshot_url, causing an UnboundLocalError that would mask the original exception. The new test correctly exercises only the post-assignment failure path, so it does not assert this buggy path — flagging for awareness, not as a PR blocker.

• Suggestion: Initialize percy_screenshot_url = '' before the try in a follow-up source PR. No action required for this test-only PR.

---

Verdict: PASS